I get a lot of customers asking about the use of ICC profiles for calibration - the calibration provided by many consumer calibration systems for PC/Mac etc.

This calibration is used for displays directly connected to the PC/Mac's graphics card, and uses ICC profiles.

Many think such calibration is equivalent to the calibration provided by LightSpace CMS and LUTs.

IT ISN'T.

Such calibration uses 1D LUTs only in a form called VCGT, and WILL ONLY match the colour temperature and gamma of a display - NOT the gamut. This means that saturation is NOT matched, and so displays will remain inaccurate.

As a result, such calibration is next to useless, but is promoted by many alternative calibration software suppliers as a valid calibration process - which it IS NOT.

Using ICC profiles within graphics systems, such as Photoshop, is a very different application, and ICC profiles generated from 3D LUTs, as with SpaceMan ICC, will be 100% accurate, including the control of Gamut (saturation).

I hope this helps with understanding the use of ICC profiles.

Steve

I'd love to see more info on the difference between the way your system's 3D LUTs work versus the ICC profile system that is used by so many in the stills world.

From my experience I have seen ICC profiles do a great job of matching saturation and Gamut. Can you fill me in more on how your system is better?

ICC profiles only work when used within a software program that understand them - Photoshop for example.

It is not possible the use ICC profiles for display calibration separately from the software program.

This alone makes them totally unusable for 'display calibration'.

This can be seen in the following two images - the first is 3D LUT calibration, the second is via a graphics card ICC profile.



The difference is very obvious - the ICC profile used within the graphics card cannot alter saturation/gamut.

Additionally, the same image, shown on the same display via 2 different software applications often looks very different, as can be seen in the following image. This has the same image displayed using the same ICC profile on two separate software programs, on the same display at the same time...



As can be seen the two images are very different... (click the image to a larger version)

This is just not accurate enough for our industry.

Hope that info helps.

Steve,

Thanks for the additional info, but I'm not completely in agreement with you on some of the details you talk about here. There are several ways ICC profiles are used to control the display of images.

For instance when it comes to Mac systems Apple's ColorSync uses the monitor profile to adjust the video card LUT controlling the monitor's display. When combined with color management savvy programs such as Adobe's Creative Suite the monitor's ICC profile works with the profile embedded in the images, (or clips), to ensure they are displayed accurately and consistently.

The bottom line is not whether an image looks better using 3D LUTs or ICC profiles, it's how "true" the display of that image is. Having extensive experience in working with images and color management I've seen over and over again how the ICC profile system can work very well.

What I am really surprised at as I learn more about how the motion side of the industry works is how far behind the still side they are when it comes to implementing even reasonable color management policies. It's almost shocking to see how haphazardly this is being done in the motion side.

Funny, I find the ICC approach to be very haphazard indeed, as can be seen by the two 'colour bar' images above.
The exact same PC screen with the exact same ICC calibration, but two different programs showing the same exact image very differently.
This shows how easily ICC calibration can fail.

This is the 'coloursync' approach you mention.
(We actually use that in Alexicc... and SpaceMan, for situations where there is no alternative).

And this still has nothing do do with the initial fact that ICC Display calibration (vcgt based) doesn't control gamut at all.

We know ICC profile workflows very well, as can be seen with SpaceMan ICC and Alexicc, so understand both approaches to a very high level.

With accurate LUT display calibration, as used in our industry, and no reliance on the ability of the 'application' to accurately understand the display, there is very little room for error. This is because the display is 'always' calibrated to a known standard.

As everyone within the industry knows these few standards - Rec709 and P3 - there really is no room for error - ever.

No PC or Mac display is ever calibrated in the same way, relying on the application to get it right - and few do unfortunately.

There are very good standards within the motion side of the industry, and they work very, very well - far better than my experience of ICC profile calibration for stills, and I come from that side of the industry historically.

This is a time where KISS really does work just as it should.

Steve,

Thanks for the additional info. If I understand your point correctly you're saying that your approach does not require the various programs to cooperate with your calibration/profiling system thus ensuring a more consistent appearance across the various programs, right?

Also your second point is that ICC profiles cannot control the gamut of the display, correct?

I'm curious how are you then controlling the appearance of the display without relying on the programs to be savvy to your color management scheme? Some programs make basic assumptions about the images or clips they are displaying and without being able to control this it's quite possible for the image to be displayed incorrectly. (This happens a lot with images viewed via browsers.)

And regarding controlling the gamut of the display my guess is that unless you have some way of actually expanding the gamut of a given device beyond what it's capable of you're talking about limiting the gamut of the display down to some "standard". For instance if my monitor has a gamut roughly equivalent to Adobe RGB (many do these days) does your system then limit it to something closer to sRGB (a la Rec709)?

If so wouldn't that compromise what I can see when I'm working with images that have a larger gamut than sRGB?

Finally what I meant about the lack of color management implementation in the motion side was that as I deal more and more with motion images it appears that ideas like embedding some sort of profile or other detailed information about the colorspace of a given clip don't seem to be broadly practiced at all.

There are various LUTs talked about, but if I get a clip from a client I'm left to figure out what LUT to use on my own. Or if a clip uses the log format I have to realize that and apply the corrections so I can see the images properly. If that information were embedded in the clips themselves and the programs were more savvy to reading this there would be a much greater consistency of appearance across the various devices and places an image is viewed on. (And correcting clips for the various destinations, web or broadcast etc, would be tremendously easier.)

I hope my questions don't come across as being too cheeky, I'm really trying to understand these issues and when I come across something like this I realize how important it is to really know what's going on beneath the more obvious level.

Indeed, this approach is universal, and requires no 'cooperation' from the systems involved - this is a very strong argument for this approach to calibration. The results are therefore easily guaranteed, and very consistent.

ICC profiles used in graphics cards cannot control gamut - just grey scale / white point.
This is because the 'vcgt' used is 1D LUT based.

All professional colour system involved in the film and tv industry are LUT capable, or use external LUT boxes, or the displays in use are themselves LUT capable - meaning that all know what the colour space in use should be - Rec709 or P3.

The negates the chance for errors, as all systems in the chain understand that the colour space is fixed.

It is impossible to expand colour gamut beyond what the display is capable of.

All displays manufactured for use in film and tv use are made to be capable of Rec709 or P3 gamut. Most 'quality' displays can be made to be Rec709 accurate via calibration via LUTs. P3 is more difficult, and is used only for high-end film work (a generalisation, but close enough).

Calibration a wide gamut display to Rec709 (sRGB basically) is easy with the calibration processes we use in this industry.

As this industry works Rec709 or P3 only, working with any 'other' wide gamut is just wrong. And there is no benefit even considering this in the real world, as all deliverables will need to be Rec709, or P3 (or maybe real celluloid film!).

As you have seen, using embedded profile based data is fraught with potential problems - using the approach we have prevents this almost totally. The main problem we have is with people that don't understand the basics of Rec709 and P3.

So no, we don't embed colour information, as we don't need to. All images are expecting to be either seen on a Rec709 or P3 calibrated display. What you then see is exactly what you should see.

All the calibration LUTs do is ensure the display you are using is correctly calibrated to either Rec709, or P3 - simples

This way consistency is guaranteed. All you need to do is ensure you display is correctly calibrated to Rec709, or maybe P3.
But as I say, P3 is really for film (cinema) based work, and is mainly used with projection as few non-projection displays can actually do P3 accurately.

Hope this helps.

Steve, I'm in the process now of trying to color grade a short film, calibrate my system in preparation, and am trying to understand how the 'moving' image side does things. I've used ICCs for a while in PS and other apps for still image workflows. My question to you, based upon your comment above, is that, like ICC, a LUT is a translation from some known state to another known state, so when you calibrate your monitor by loading a LUT, you must be assuming an initial known state, right? For example, if you calibrate your monitor (load a LUT) to Rec709, then I assume what you are actually doing is creating a LUT for the monitor to 'correct' an incoming Rec 709 signal to display as a true Rec 709 image on that monitor. Whereas, if I understand this correctly, an output ICC profile maps a known working baseline (like CIEXYZ) to a particular hardware device like your monitor. If this is true, then for monitor calibration in particular, the difference would appear to be conversion in the case of ICC and 'correcting' in the case of LUTs (I know LUTs are used for conversion too, like sLog to Rec 709 for ingest, but I'm talking monitor calibration here). Given the above, would not the ICC profiles work the same if you were outputting an Rec 709 signal when you calibrated your monitor, creating the ICC?

In my particular case, I have a Dell U3011 which is a 30" 10-bit display, connected to a nVidia Quadro 4000 via displayport. The monitor has an sRGB preset as well as an AdobeRGB preset, along with manual controls (which can be used for calibration). Assuming the sRGB preset was correct (Rec 709, right?), and I was using Adobe Speedgrade (for example) to color grade, would I be seeing accurate colors? If not, why not?

Additionally, if my ultimate goal was theatrical distribution, but my display is not P3 capable, is there a way to take advantage of the wider AdobeRGB gamut of the display in the grading workflow (display LUTs in SG, etc.), and should I even consider that given the source is H.264 4:2;0 8-bit to begin with?

Would appreciate any advice/clarification.

Cheers,
Griz

The issue is no display preset - especially those in a relatively cheap display such as a Dell - can be assumed to be accurate.
They will not be.
You NEVER assume anything when looking for accurate calibration.

You must profile the display and generate a calibration LUT from that profile.

For theatrical delivery you can use the maximum gamut of any display for grading, so long as you have fully profiled the display and then generate a LUT to convert from that profile to P3 for the delivery.
(And very few displays can actually accurately cover AdobeRGB)

Steve

Steve, thanks for the response. The Dell in question, the U3011, is not that cheap, at just under $2k. Although it does no have HW LUT capability, and the presets have been shown to be not very accurate, proper calibration with xRite iPro has shown very accurate results. Also, it has been shown to reproduce 116% of the AdobeRGB color space.


In my case, what I do is use the i1Pro to profile the monitor, creating an ICC profile that is then loaded into the Quadro 4000 card as a SW calibration.


I'm not sure I understand what you mean. So, if I, say, profile my monitor to display the AdobeRGB color space accurately, color grade with SpeedGrade (using what LUT settings, if any), then would need another conversion from AdobeRGB to P3 when outputting my DCP? I do see built-in LUTs that might include Rec 709 to P3, so is that why people rely on Rec 709 (because there are transforms for them already)? Am I on the right track, or am I just missing it completely?

Cheers

Have you used LightSpace with a decent probe to verify the calibration?
As yet, we have never found a manufacturer calibration system that actually works...

Unfortunately, using X-Rite's 'Profiler' software is not going to be accurate.

And as has been said, if you use a professional grading system they will not be ICC compliant, so and ICC based colour management will be inaccurate, as that will just use the 1D LUT (VCGT) component of the ICC profile.

And yes, if you use the display in Wide Gamut mode (any default setting will not be accurate) you will need to generate a LUT to convert from the Display's native gamut and gamma to the target DCI. You cannot use a generic AdobeRGB setting, as that will not be accurate.

As I said, you cannot rely on any presets, as they will not be accurate enough.

Steve

I have not used LightSpace, but have downloaded the free DPS version to try. Your site has confused me a bit on the difference between the terms "Calibration" and "Profiling". Can you elaborate on the difference please, in the context of CM in general?

My question then is, if the DPS version can really do manual calibration, could I then use it to calibrate my display for correct use with the equipment described above (i.e., Quadro 4000, Dell U3011 over DisplayPort, and my i1Pro Spectrometer)? If so, once I calibrate/profile (whichever is the correct term - see above), how do I ensure it is set up properly, and used correctly by my applications (Adobe Speedgrade / Premiere Pro / AE)?

I appreciate your continued support.

Shane

Profiling is the measurement of the display's native response.

Calibration is the generation of a 3D LUT to match the display to a specific colour space standard.

And no, manual calibration will just get a display to be 'close' to an average level of calibration.
That will never be 'calibrated'.

To actually calibrate you will need to control the volumetric colour space, and manual calibration just moves the basic colour points - average greu scale, saturation, and luminance - not the total colourmetric colour accuracy.

To perform full volumetric calibration you need a way to use a 3D LUT, such as a display with 3D LUT capability (Eizo, FSI), or an external LUT box (AJA, eeColour, IS-mini, BoxIO), or a software system that uses 3D LUTs on the output (Resolve, Mistika, Scratch, etc).

Steve

I think I've got it. Thanks Steve.

Steve,

I'm a little confused about how to adjust my monitor controls on my Dell U3011 before/during calibration. At the moment, just trying to wrap my head around this, I'm using the xrite i1Profiler because I already own it and am somewhat familiar with it, along with the i1Pro spectrophotometer. In it, I can choose a Rec BT709 'workflow', and it sets defaults for me like D65 and 80 cd/m3 and gamma of 2.2, but I also have options for Native, xy, and Measure (along with the standard D50, D55, etc.) for White Point. The contrast ratio can be set to one of the following: Native, iCC PCS Black Point (287:1), Custom Black Point, Measured Black Point, or Custom which allows me to set a value between 10 and 1000, but I don't know what I should use as a target. The White point is specified as x=0.313 and y=0.329. I also have the option of Matrix based or Table based profiles produced, several Chromatic Adaptation methods, and ICC Ver 2 and 4.

I can tell it to let me manually adjust Contrast, Brightness, and RGB Controls. On my monitor OSD menu, I have a control for Brightness and Contrast, but not sure what I should set these too to begin with. Also, along with a bunch of presets like AdobeRGB, sRGB, Movie, Standard, etc., I have Custom Color, which then allows me to adjust any and all of the following:

Gain for R, G, B
Offset for R, G, B
Hue for R, G, B, C, M, Y
Saturation for R, G, B, C, M, Y

While profiling, only Gain and Offset in the Custom Color settings change anything. Changing Hue and/or Saturation seem to have zero affect (while profiling in i1Profiler - it does affect things outside of this app). Likewise, I can adjust Brightness when it asks me to match my target of 80 cd/m3, but it never asks me to adjust Contrast (which was set randomly to a previous profile). In the past, I've read that you turn Contrast all the way up before profiling, but that really washes things out, even when I turn down the Brightness, so I'm at a loss as to how to configure these monitor settings at the start of profiling.

I'm able to adjust the Gain and Offset on the monitor to get nearly exactly the x,y and K of the target, with Gains at 98-100 for R, G, B, and Offsets between 30 and 50 for R, G, B (range 0 - 100). I don't know how the Gain/Offset interact with the Brightness/Contrast controls, since it seems to be over-specified in that regard.

I can also create a profile for each of my two monitors separately, and they are each loaded into my Quadro 4000 correctly for each monitor, so I know it has dual LUT capability, since I can change them independently. Surprisingly, 80 cd/m3 still seems really bright to me (seated in a darkend room), and definitely brighter than I had it before I calibrated. I've also noticed something (perhaps metamerism?) caused by (I suppose) the LCD backlight when viewing different parts of the screen that seems to reduce contrast.

These monitor controls questions/issues will be evident for me with any CMS I use, so I'm just trying to get a handle on them first. Any insight/suggestions on settings would be GREATLY appreciated.

Thanks,
Shane